Sevda ÖĞÜT

Excited to share CellBridge, our new preprint led by @sevdoche in collaboration with @pafrossard and @DorinaThanou at @EPFL_en! 🔬 Looking at an H&E slide, we can learn a remarkable amount about a tissue: its architecture, the cells within it, and how they interact. But identifying exactly which cells are present can be harder. A lymphocyte may be recognisable, while telling whether it is a T cell or a B cell often requires additional markers. Spatial transcriptomics provides that molecular detail by measuring gene expression in individual cells. But it is much harder to obtain routinely than an H&E slide. Could we use these molecular measurements to teach an image model more about cell identity and help it recognise when appearance alone is not enough? This is the idea behind CellBridge. We trained CellBridge on millions of cells with paired H&E images and Xenium spatial transcriptomic profiles. During training, the model learns to connect each cell’s appearance with its molecular identity. Once trained, it predicts cell types from H&E images alone. Across our benchmarks, CellBridge improved fine-grained cell classification over leading pathology models, with gains of up to 16 percentage points in macro F1. But better classification is only part of the story. Some biologically important distinctions remain difficult to resolve from an image. Our Cell-Ontology-Fallback (COF) module lets the model adjust how specific its answer should be. For example, when the evidence does not support distinguishing a T cell from a B cell, it can report “lymphocyte” rather than commit to an uncertain subtype. COF combines the biological relationships in the Cell Ontology with branch-calibrated conformal prediction to make these decisions. What excites me is the possibility of extracting more cellular detail from routinely available slides while making clearer where additional molecular measurements are still needed. 🎥 Our 1-minute video walks through the idea. Jointly supervised by @pafrossard, @DorinaThanou, and me. A huge thank you to our co-authors: @arpitdm, @AmirhVahidi, Dabin Jeong, @SebastianBirk_, Adam Boxall, @Adib_m_, Marta Rosa Sallese, and Yinshui Chang. 📄 Preprint: biorxiv.org/content/10.64898… 💻 Code: github.com/Lotfollahi-lab/ce… 📓 Tutorial: github.com/Lotfollahi-lab/ce… #ComputationalPathology #SpatialOmics #CellTypeAnnotation #MultimodalLearnin

Excited to share TERRA, a tissue world model 🧬 Over ~1.5 years we ran a large data-generation + modelling effort to build a world model for human tissues, pretrained on 112M cells from spatial transcriptomics (mostly Xenium 5000-plex + public data). It's built on one of the largest human spatial transcriptomics corpora assembled to date, spanning 20 tissues across development, health and 26 disease conditions, ~two-thirds newly generated in-house. Why a "world model" for tissue? Images have universal representations (ViT/DINOv3), so do proteins (ESM, @alexrives) and pathology (UNI, @AI4Pathology). We've worked hard to build something similar for human tissue: one model that captures its multi-scale logic, genes → cells → their native microenvironments. Like the JEPA approach @ylecun has championed, TERRA learns by prediction in embedding space, but for human tissue. How it works: it tokenises each cell together with its nearest neighbours into one sequence while keeping every gene's identity, then masks part of a neighbourhood and predicts the representation of the hidden part, not raw noisy counts. From one backbone it reads out three scales, gene embeddings (what a gene is doing in a cell and its niche), cell embeddings (cell type and state) and neighbourhood embeddings (the niche), and because it keeps gene-level resolution it can knock a gene out in silico and predict the response. Applied entirely zero-shot, TERRA maps and perturbs human tissue across unseen organs, diseases and technologies, outperforming existing spatial approaches. Three take-homes: 1️⃣ One model, any tissue. A single pretrained backbone provides tissue representations zero-shot, handling genes, cells and niches across organs and platforms, off the shelf. 2️⃣ New biology, development to clinic. We built a new spatial atlas of the developing human pancreas and found an islet-associated capillary state that looks like a precursor of mature islet vasculature. In kidney, TERRA's in silico knockouts predicted the tissue-injury programme from cancer immunotherapy (checkpoint blockade), confirmed in treated kidneys, detected in blood, and linked to declining kidney function. 3️⃣ A grammar of tissue architecture. By coupling each cell's state to its niche, TERRA defines recurring cross-organ "archetypes" of macrophage neighbourhoods, including a tumour-boundary niche that tracks poor survival in kidney cancer. TERRA is already in use: it powered our recent skin atlas of hidden immune-memory niches (biorxiv.org/content/10.64898…), with more studies coming soon. This was an amazing collaboration between clinicians, machine-learning scientists and cell biologists 🙏 Led by @SebastianBirk_, @ValiSanian @AmirhVahidi, Samuel Ogden, @daniyal_jafree, @Adib_m_, @CarloLeonardi7 and Arpit Merchant, with Lassi Paavolainen, Menna Clatworthy, @bayraktar_lab, @Muzz_Haniffa, Tom Mitchell and @bakhti_mostafa. Huge thanks too to everyone who shared data and helped along the way. What excites me most is seeing how the community builds on this. The model, code and tutorials are all public, so anyone can run TERRA on their own tissues, extend it, or build new models on top. Huge thanks to the whole team across @sangerinstitute and our many collaborators. 📄 Paper: biorxiv.org/content/10.64898… 💻 Code: github.com/Lotfollahi-lab/te… 🤗 Model: huggingface.co/lotfollahi-la… #SpatialTranscriptomics #SpatialGenomics #FoundationModels #AI4Science #MachineLearning #ComputationalBiology #SingleCell #WorldModels